Every time I write about artificial intelligence, I struggle with which image to use. The obligatory blue mosaic face, or Matrix-style falling code? Maybe a glowing brain with neurons in the background, or a white humanoid robot with a touchscreen? All of these visuals are not only tiresomely overused clichés – according to the “Better Images of AI” study, among others, they are also dangerously misleading.
The researchers Dr Kanta Dihal and Tania Duarte point out that these standard depictions of AI create false expectations, spread fear of threats that do not exist, and at the same time draw attention away from the real problems.
As one participant in their research put it rather well: “AI is always scary and always separates us from the people we are trying to communicate with… The visuals simply do not serve their purpose in the sphere where we use them.”
So what should we do about it? The authors of the study offer a useful framework for depicting AI responsibly, built on four principles: honesty (show what the technology actually does), humanity (show people working with AI, not AI as a person), necessity (do not use AI imagery when the subject does not call for it) and specificity (match the image to the particular technology).
And what instead of those glowing blue brains? Try showing the real settings where AI is used. Show a variety of people working with the technology. Capture the physical infrastructure that makes AI possible – servers, cabling, cooling systems.
This is not merely a question of aesthetics. The way we present AI visually directly shapes how the public understands it and how people relate to it. At a time when AI increasingly touches our lives, depicting the technology accurately is essential for informed public decisions.
So next time you are picking an image for your post about artificial intelligence, forget the Terminator. Your content – and our collective perception of the technology – deserves better.
The document in brief:
What to avoid:
1. The colour blue – associated with technological progress, but also perceived as cold, clinical and remote
2. Falling code – a nod to The Matrix, which builds a dystopian idea of AI
3. The human brain – a misleading link between AI and human intelligence that overstates what the technology can do
4. Science-fiction references – the Terminator or 2001: A Space Odyssey – and anthropomorphism, which create unrealistic expectations
Best practices:
1. Show what AI can really do, without exaggeration
2. Show people working with AI technologies, not AI as a being
3. Use specific imagery for specific kinds of AI technology (biometric systems, image recognition, classification)
4. Show the material reality of AI, hardware and data centres included
5. Capture AI as it is now, not a fantastical future
6. Build metaphors that capture the nature of AI better than a “brain” or a “thinking machine”
Illustration: Janet Turra & Cambridge Diversity Fund / https://betterimagesofai.org
